[c-nsp] HSRP vs VRRP

[Rubens Kuhl Jr.]

> > Switch failure is a different matter. The hosts would need to be
> > multi-homed to survive that.
>
> *And* need to run a dynamic routing protocol to figure out which
> network is still usable and which default gateway to use (they still
> might have a local link, depending on the way the switch fails)

Using failover at the NIC driver level can make this easier(both
Broadcom and Intel drivers have such features). There is some kind of
protocol involved (either a connection monitor with active/standby, or
spanning-tree with active/active), but deployment is somewhat easier
than extending an IGP to the hosts.

I say "somewhat easier" because there are still those nasty
firmware-versions x driver-versions x OS-versions x network-behaviour
x failure-modes matrix where some of the possibilities won't work, but
who said IT is easy ?

Rubens